Learning Words Actively in Context: A Quasi-Naturalistic Hyperscanning Study

• In Event: Why Parent-Infant Interaction Matters: Impacts of Parent’s Multimodal Inputs on Infant’s Perceptual Experiences

Abstract

Recent research has shown that infants actively and strategically sample information from their environment as a function of their uncertainty (Baer & Kidd, 2022; Goupil & Kouider, 2019), and that they learn novel properties or labels for objects better after choosing to visually explore these objects (Begus, Southgate, & Gliga, 2015), or pointing towards them (Begus & Southgate, 2012; Lucca & Wilbourn, 2016). Yet, so far, evidence for active learning during infancy comes from highly structured experiments where information is consistently delivered at specific times and locations by the experimenters. This greatly simplifies the task for the infant, as compared to what they typically encounter in naturalistic, complex and free-flowing situations. Whether and how caregiver-infant dyads self-organize to support infant-led forms of learning early on during naturalistic interactions remains, therefore, unknown

Adopting a dyadic approach, in this pre-registered study we collected audio-visual recordings and dual-EEG data from 40 caregiver-infant dyads while infants (14-mo, an addition 23 dyads participated but did not contribute usable data) were provided with the opportunity to learn novel words from their caregivers in a naturalistic setting (Fig. 1). Based on parental reports, we selected objects that are either familiar (e.g., dog) or totally novel (e.g., pliers) for the child. Dyads briefly played with the toys, and infants’ learning of new object-label mappings was then tested in a looking-while-listening procedure (Fernald & Marchman, 2008). This setting allows us to examine what fosters word learning in these (quasi-)naturalistic conditions, and whether infants’ active engagement with objects helps them learn novel words.

We now summarize preliminary findings with the subsample of data that is already pre-processed (N=31, data for the additional 9 dyads is currently being preprocessed/coded). Results show that infants learn what novel words refer to in this context (t(30) = 3.2, p = 0.003), with some variability. Consistent with previous reports (Wass, Clackson, et al., 2018; Yu & Smith, 2013), we observe that – rather than mutually adapting to one another’s behaviors, i.e., coordinating their gaze in the strict sense – 14-mo and their mothers engage in a very asymmetric form of interaction during play. While infants often proactively initiated looks towards objects, and rarely followed their mother’s gaze, mothers looked towards their infants’ face most of the time, and tended to follow their interests (Fig.2A). Results also suggest that infants learn the name of a novel object better when their caregiver named the object more often after they proactively initiated a look towards that object (Fig 2.B, t(30) = 1.94, p = 0.057).

Regarding brain data, preliminary analyses suggest that infants show a stronger Nc ERP component when they lead a look towards an object, as compared to following their mothers’ attention towards an object. Further analyses will focus on examining time-frequency responses and cross-dyadic brain measures (Marriott Haresign et al., 2022) around gaze shifts and naming onsets, in order to elucidate further how each partner’s attention is distributed during naturalistic play, and how this relates to infant learning.

Authors

• Louise Goupil, Centre National de la Recherche Scientifique (CNRS)
  Presenting Author

• Katherine Denman, University of East London

• Zion Henry

• Victor Serveau

• Ira Marriott Haresign, University of East London

• Isabelle Dautriche

• Sam V Wass, University of East London